Decentralized multi-agent entropy-driven exploration under sparsity constraints

This paper proposes a new algorithm, which uses the second order information of a Least Absolute Shrinkage and Selection Operator (LASSO) to achieve an active sensing approach driven by minimizing the entropy of sparse unknown environments, for the multi agent case. For this, a signal model, which restricts the agent's measurements according to its sensor's view, is introduced into the Distributed LASSO (DLASSO) framework. With the help of Compressed Sensing (CS), the DLASSO is able to estimate the environment with less measurements. After the DLASSO converged to a solution, each agent evaluates the proposed algorithm for choosing new measurement locations.